mirror of
https://github.com/openai/codex.git
synced 2026-11-05 20:10:33 +00:00
fix(core): align session title prompt with app
Use the same title generation instructions and length limits as the Codex App so generated session names behave consistently across clients. Normalize generated names with the app constraints before persisting them.
This commit is contained in:
Felipe Coury
+83
-41
@@ -446,21 +446,42 @@ pub(crate) const SUBMISSION_CHANNEL_CAPACITY: usize = 512;
|
||||
const CYBER_VERIFY_URL: &str = "https://chatgpt.com/cyber";
|
||||
const CYBER_SAFETY_URL: &str = "https://developers.openai.com/codex/concepts/cyber-safety";
|
||||
const DEFAULT_THREAD_TITLE_MODEL: &str = "gpt-5.4-mini";
|
||||
const THREAD_TITLE_PROMPT: &str = r#"Create a short activity label for this Codex session.
|
||||
|
||||
const THREAD_TITLE_INPUT_CHAR_LIMIT: usize = 2_000;
|
||||
const THREAD_TITLE_MIN_CHARS: usize = 18;
|
||||
const THREAD_TITLE_MAX_CHARS: usize = 36;
|
||||
const THREAD_TITLE_PROMPT: &str = r#"You are a helpful assistant. You will be presented with a user prompt, and your job is to provide a short title for a task that will be created from that prompt.
|
||||
The tasks typically have to do with coding-related tasks, for example requests for bug fixes or questions about a codebase. The title you generate will be shown in the UI to represent the prompt.
|
||||
Generate a concise UI title (18-36 characters) for this task.
|
||||
Return JSON with exactly one field: {"title": "..."}.
|
||||
The title must be at least 2 words and at most 6 words, 48 characters or fewer, no markdown, no surrounding quotes, no trailing period.
|
||||
Name the user's concrete intent, not the wording of the prompt.
|
||||
Mention the most useful differentiating entity when one exists: a PR number, file, command, feature, error, ticket, or product name.
|
||||
Do not force paths, project names, or branch names when a clear plain-language task is better.
|
||||
The title value must be plain text. No quotes or trailing punctuation.
|
||||
Do not use markdown or formatting characters.
|
||||
If the task includes a ticket reference (e.g. ABC-123), include it verbatim.
|
||||
|
||||
Style:
|
||||
- Write a compact activity label, not a document heading.
|
||||
- Prefer sentence case: capitalize only the first word, proper nouns, acronyms such as PR/API/MCP/SSH, commands, file paths, ticket IDs, and product names.
|
||||
- Prefer an -ing verb when it sounds natural: "Reviewing PR description", "Investigating startup regression", "Checking branch purpose".
|
||||
- Use simple familiar verbs. Avoid uncommon verbs like "inventorying"; use "Researching statusline options" or "Listing statusline options" instead.
|
||||
- Use an imperative verb only when it is the natural command/task wording: "Rebase and resolve conflicts", "Install timeout globally", "Run $documentation-pass".
|
||||
- Avoid title case like "Branch Purpose Inquiry" or "Review PR Description"."#;
|
||||
Generate a clear, informative task title based solely on the prompt provided. Follow the rules below to ensure consistency, readability, and usefulness.
|
||||
|
||||
How to write a good title:
|
||||
Generate a single-line title that captures the question or core change requested. The title should be easy to scan and useful in changelogs or review queues.
|
||||
- Use an imperative verb first: "Add", "Fix", "Update", "Refactor", "Remove", "Locate", "Find", etc.
|
||||
- Aim for 18-36 characters; keep under 5 words where possible.
|
||||
- Capitalize only the first word unless locale requires otherwise.
|
||||
- Write the title in the user's locale.
|
||||
- Do not use punctuation at the end.
|
||||
- Output the title as plain text with no surrounding quotes or backticks.
|
||||
- Use precise, non-redundant language.
|
||||
- Translate fixed phrases into the user's locale (e.g., "Fix bug" -> "Corrige el error" in Spanish-ES), but leave code terms in English unless a widely adopted translation exists.
|
||||
- If the user provides a title explicitly, reuse it (translated if needed) and skip generation logic.
|
||||
- Make it clear when the user is requesting changes (use verbs like "Fix", "Add", etc) vs asking a question (use verbs like "Find", "Locate", "Count").
|
||||
- Do NOT respond to the user, answer questions, or attempt to solve the problem; just write a title that can represent the user's query.
|
||||
|
||||
Examples:
|
||||
- User: "Can we add dark-mode support to the settings page?" -> Add dark-mode support
|
||||
- User: "Fehlerbehebung: Beim Anmelden erscheint 500." (de-DE) -> Login-Fehler 500 beheben
|
||||
- User: "Refactoriser le composant sidebar pour reduire le code duplique." (fr-FR) -> Refactoriser composant sidebar
|
||||
- User: "How do I fix our login bug?" -> Troubleshoot login bug
|
||||
- User: "Where in the codebase is foo_bar created" -> Locate foo_bar
|
||||
- User: "what's 2+2" -> Calculate 2+2
|
||||
|
||||
By following these conventions, your titles will be readable, changelog-friendly, and helpful to both users and downstream tools."#;
|
||||
|
||||
impl Codex {
|
||||
/// Spawn a new [`Codex`] and initialize the session.
|
||||
@@ -4118,8 +4139,10 @@ impl Session {
|
||||
let sess = Arc::clone(self);
|
||||
let title_request = ThreadNameGenerationRequest {
|
||||
sub_id: turn_context.sub_id.clone(),
|
||||
user_message: truncate_string_chars(&initial_user_message, 4_000),
|
||||
cwd: turn_context.cwd.to_path_buf(),
|
||||
user_message: truncate_string_chars(
|
||||
&initial_user_message,
|
||||
THREAD_TITLE_INPUT_CHAR_LIMIT,
|
||||
),
|
||||
model_name: turn_context
|
||||
.config
|
||||
.session_titles
|
||||
@@ -5068,7 +5091,6 @@ struct ThreadNameOutput {
|
||||
struct ThreadNameGenerationRequest {
|
||||
sub_id: String,
|
||||
user_message: String,
|
||||
cwd: PathBuf,
|
||||
model_name: String,
|
||||
model_reasoning_summary: ReasoningSummaryConfig,
|
||||
service_tier: Option<ServiceTier>,
|
||||
@@ -5117,7 +5139,7 @@ async fn generate_thread_name(
|
||||
id: None,
|
||||
role: "user".to_string(),
|
||||
content: vec![ContentItem::InputText {
|
||||
text: build_thread_name_input(request.cwd.as_path(), &request.user_message),
|
||||
text: format!("User prompt:\n{}\n", request.user_message.trim()),
|
||||
}],
|
||||
end_turn: None,
|
||||
phase: None,
|
||||
@@ -5170,40 +5192,60 @@ fn thread_name_output_schema() -> Value {
|
||||
serde_json::json!({
|
||||
"type": "object",
|
||||
"properties": {
|
||||
"title": { "type": "string" }
|
||||
"title": {
|
||||
"type": "string",
|
||||
"minLength": THREAD_TITLE_MIN_CHARS,
|
||||
"maxLength": THREAD_TITLE_MAX_CHARS,
|
||||
}
|
||||
},
|
||||
"required": ["title"],
|
||||
"additionalProperties": false
|
||||
})
|
||||
}
|
||||
|
||||
fn build_thread_name_input(cwd: &Path, user_message: &str) -> String {
|
||||
let git_root = get_git_repo_root(cwd);
|
||||
let project = git_root
|
||||
.as_deref()
|
||||
.and_then(Path::file_name)
|
||||
.or_else(|| cwd.file_name())
|
||||
.map(|name| name.to_string_lossy().to_string());
|
||||
match project {
|
||||
Some(project) => format!(
|
||||
"Project: {project}\n\nUser's first message:\n{}\n",
|
||||
user_message.trim()
|
||||
),
|
||||
None => format!("User's first message:\n{}\n", user_message.trim()),
|
||||
}
|
||||
}
|
||||
|
||||
fn sanitize_generated_thread_name(name: &str) -> Option<String> {
|
||||
let collapsed = name.split_whitespace().collect::<Vec<_>>().join(" ");
|
||||
let stripped = collapsed
|
||||
let first_line = name
|
||||
.replace("\r\n", "\n")
|
||||
.lines()
|
||||
.find(|line| !line.trim().is_empty())?
|
||||
.trim()
|
||||
.trim_matches(['"', '\'', '`', '*', '_'])
|
||||
.trim_end_matches('.')
|
||||
.trim();
|
||||
if stripped.is_empty() {
|
||||
.to_string();
|
||||
let prefixed = first_line.trim();
|
||||
let stripped_title = if prefixed
|
||||
.get(.."title".len())
|
||||
.is_some_and(|prefix| prefix.eq_ignore_ascii_case("title"))
|
||||
&& prefixed["title".len()..]
|
||||
.chars()
|
||||
.next()
|
||||
.is_some_and(|separator| separator == ':' || separator.is_whitespace())
|
||||
{
|
||||
prefixed["title".len()..].trim_start_matches(|ch: char| ch == ':' || ch.is_whitespace())
|
||||
} else {
|
||||
prefixed
|
||||
};
|
||||
let stripped = stripped_title
|
||||
.trim_matches(['"', '\'', '`', '“', '”', '‘', '’'])
|
||||
.split_whitespace()
|
||||
.collect::<Vec<_>>()
|
||||
.join(" ")
|
||||
.trim_end_matches(['.', '?', '!'])
|
||||
.trim()
|
||||
.to_string();
|
||||
let stripped = stripped.as_str();
|
||||
let char_count = stripped.chars().count();
|
||||
if char_count < THREAD_TITLE_MIN_CHARS {
|
||||
return None;
|
||||
}
|
||||
Some(truncate_string_chars(stripped, 48))
|
||||
if char_count > THREAD_TITLE_MAX_CHARS {
|
||||
let truncated = stripped
|
||||
.chars()
|
||||
.take(THREAD_TITLE_MAX_CHARS - 1)
|
||||
.collect::<String>()
|
||||
.trim_end()
|
||||
.to_string();
|
||||
return Some(format!("{truncated}…"));
|
||||
}
|
||||
Some(stripped.to_string())
|
||||
}
|
||||
|
||||
fn truncate_string_chars(value: &str, max_chars: usize) -> String {
|
||||
|
||||
@@ -162,8 +162,21 @@ fn generated_thread_names_are_sanitized_for_display() {
|
||||
);
|
||||
assert_eq!(
|
||||
sanitize_generated_thread_name("Fix TypeScript\nschema\tgeneration").as_deref(),
|
||||
Some("Fix TypeScript schema generation")
|
||||
None
|
||||
);
|
||||
assert_eq!(
|
||||
sanitize_generated_thread_name("TITLE Locate foo_bar creation?").as_deref(),
|
||||
Some("Locate foo_bar creation")
|
||||
);
|
||||
assert_eq!(
|
||||
sanitize_generated_thread_name("Title: Locate foo_bar creation?").as_deref(),
|
||||
Some("Locate foo_bar creation")
|
||||
);
|
||||
assert_eq!(
|
||||
sanitize_generated_thread_name("Refactor an especially noisy sidebar component").as_deref(),
|
||||
Some("Refactor an especially noisy sideba…")
|
||||
);
|
||||
assert_eq!(sanitize_generated_thread_name("Fix bug"), None);
|
||||
assert_eq!(sanitize_generated_thread_name(" \n\t "), None);
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user